加速器磁铁电源的被控对象辨识模块设计

Plant identification module design for accelerator magnet power supplies

国内加速器磁铁电源的控制策略基本上以比例-积分-微分(Proportion-Integral-Derivative,PID)控制为主,对被控对象的处理是基于物理模型归纳传递函数,并在此基础上设计控制器。该方法受制于物理模型中元器件参数的不精确性和结构上的不确定性,更重要的是控制器设计一般不关注被控对象的内部机理而是其输入输出(Input-output,I/O)特性。本文提出一种可工作于电源数控平台上的辨识建模方法,其表现出了更好的实时性、适用性和通用性。借助一类子空间模型辨识方法(Multivariable Output Error State s Pace,MOESP),在现场可编程逻辑门阵列(Field Programmable Gate Array,FPGA)上有针对性地设计模块,并使用可编程片上系统(System On a Programmable Chip,SOPC)集成软核处理器来完成数据处理和参数计算。基于北京正负电子对撞机II期(Beijing Electron Positron Collider II,BEPCII)和加速器驱动次临界系统(Accelerator Driven Sub-critical System,ADS)磁铁电源数控平台的被控对象辨识模块已成功运用在BEPCII和ADS电源样机上。经过严格的测试表明,辨识模型可以对电源的实际输出电流作出较精确的预报。该辨识模块易于使用,为控制系统设计提供了关键信息,适用于各种特性的负载。

Background： The control strategy of accelerator power supplies mainly depends on PID （Proportion- integral-derivative） controlling at domestic plant. The controlled plant is treated as transfer fimctions induced from physical models and the controller design depends on them. This approach suffers from the shiffing between design values and the real elements as well as the uncertainty of the hardware structure. Moreover, the engineers are mainly not interested in the internal mechanisms of the plants but their input-output （I/O） behavior. Purpose： This study aims to design a plaxtt identification module with better real-time performance, applicability and versatility. Methods： Based on subspace model identification, particularly the MOESP （Multivariable Output Error State sPace） method, the FPGA （Field Programmable Gate Array） modules are designed in pertinence and the identification algorithm is processed by embedded SOPC （System On a Programmable Chip）. These modules were applied to magnet power supply digital control platform for both BEPCII （Beijing Electron Positron Collider II） and ADS （Accelerator Driven Sub-critical System）. Results： The identified model was strictly tested and proved to be capable to predict the output current with significant accuracy for magnet power supplies of both BEPCII and ADS. Conclusion： The module is easy to use for providing key information for controller design and compatible with loadings of various characteristics Compared with traditional analytic modelling, the plant identification module performs better in applicability, versatility and real-time performance.